Face Mask Recognition and Normative Detection Algorithm Based on Attention Mechanism Fusion and Image Filtering

Under the current normalized management of the epidemic situation, due to the dense population in public places and small detection targets, it is difficult for manual supervision to play a certain role. At the same time, the existing recognition algorithms still have certain deficiencies in whether masks are worn or not. Face mask recognition based on YOLOv5s and fusion of multi-scale attention features and wearing standard detection algorithm using HSV and YCrCb image filtering. The experimental results show that after the introduction of the CBAM attention module, the mAP reaches 73.76%, which is an increase of 1.69%. The skin color of the mouth and nose area of the face is extracted, and the accuracy of mask wearing standard detection reaches 84.93%, which meets the needs of practical applications.